Real-Time Model-Free Detection of Low-Quality Synchrophasor Data Meng Wu and Le Xie Department of Electrical and Computer Engineering Texas A&M University College Station, TX NASPI Work Group meeting March 23, 2017 1
Presentation Outline 1 Introduction 2 Technical Approach 3 Case Studies 4 Conclusions & Future Work 2
Motivation: PMU Data Quality Problems Current Practice PMU-based decision making tools require accurate PMU data for reliable analysis. PMU data has higher sampling rate and accuracy requirement. Typical PMU bad data ratio in California ISO ranges from 10% to 17% (in 2011) [1]. Critical Needs Urgent need to develop scalable, real-time methods to monitor and improve PMU data quality. Conventional bad data detection algorithms are rendered ineffective, novel algorithms are needed. [1] California ISO, Five year synchrophasor plan, California ISO, Tech. Rep., Nov 2011. 3
Current Approaches for PMU Bad Data Detection PMU-based state estimator [2]. Kalman-filter-based approach [3]. Require system parameter and topology information. Require converged state estimation results. Low-rank matrix factorization for PMU bad data detection [4]. Pre-defined logics & thresholds for bad data detection [5]. Matrix factorization involves high computational burden. Robustness of pre-defined logics under eventful conditions. [2] S. Ghiocel, J. Chow, et al. Phasor-measurement-based state estimation for synchrophasor data quality improvement and power transfer interface monitoring. [3] K. D. Jones, A. Pal, and J. S. Thorp, Methodology for performing synchrophasor data conditioning and validation. [4] M. Wang, J. Chow, et al. A low-rank matrix approach for the analysis of large amounts of power system synchrophasor data. [5] K. Martin, Synchrophasor data diagnostics: detection & resolution of data problems for operations and analysis. 4
Overview of The Proposed Approach [6] Problem Formulation Study spatio-temporal correlations among good / eventful / bad PMU data. Formulate bad PMU data as spatio-temporal outliers among other data. Apply density-based outlier detection technique to detect bad PMU data. Online PMU Bad Data Detection Algorithm [6] M. Wu, and Le Xie. Online Detection of Low-Quality Synchrophasor Measurements: A Data-Driven Approach. IEEE Transactions on Power Systems. Accepted, to appear. 5
Presentation Outline 1 Introduction 2 Technical Approach 3 Case Studies 4 Conclusions & Future Work 6
Good Data VS Eventful Data VS Bad Data Phase Angle Measured by A Western System PMU for A Recent Brake Test Event Event Bad Data Bad Data Bad Data Bad Data Weak Temporal Correlation Weak Temporal Correlation Weak Temporal Correlation Weak Temporal Correlation Weak Temporal Correlation 7
Good Data VS Eventful Data VS Bad Data Event Bad Data Bad Data Bad Data Bad Data Strong Spatial Correlation Weak Spatial Correlation Weak Spatial Correlation Weak Spatial Correlation Weak Spatial Correlation 8
Features of Good / Eventful / Bad Data Criteria: Good Data VS Eventful Data VS Bad Data Good Data: strong spatio-temporal correlations with its neighbors. Eventful Data: weak temporal but strong spatial correlations with its neighbors. Bad Data: weak spatio-temporal correlations with its neighbors. PMU Bad Data: Spatio-Temporal Outlier 9
Example of Spatio-Temporal Correlations Voltage Magnitude (pu) 1.02 1 0.98 0.96 0.94 0 1 2 3 4 5 6 7 8 Time (s) 8 PMU curves 3 time windows 2 instants / window Map 3 8 Curves To 2D Euclidian Space 8 green points (normal window) 8 blue points (fault-on window) 8 red points (low-quality window) 2 red outliers (low-quality data) 10
Quantification of Spatio-Temporal Correlations [6] Definition of Normalized Standard Deviation Normalized standard deviation: Spatio-Temporal Correlation Metrics (Distance Function) For high-variance bad data: High-variance bad data: data spikes, data loss, high noise, false data injections, etc. Explanation: Standard deviation of PMU curve obtained from i th PMU channel at k th time window, normalized by the average standard deviation of the historical clean data of the same PMU channel. For low-variance bad data: Low-variance bad data: un-updated data, etc. [6] M. Wu, and Le Xie. Online Detection of Low-Quality Synchrophasor Measurements: A Data-Driven Approach. IEEE Transactions on Power Systems. Accepted, to appear. 11
Online Detection of Low-Quality PMU Data [7] Spatio-Temporal Correlation Metrics (Distance Function) For high-variance bad data: Density-Based Local Outlier Detection Local Reachability Density: High-variance bad data: data spikes, data loss, high noise, false data injections, etc. Local Outlier Factor [12]: For low-variance bad data: Bad Data Detection: Low-variance bad data: un-updated data, etc. LOF(p) >> 1: p contains bad data. LOF(p) 1: p contains good data only. [7] Breunig, Markus M., et al. "LOF: identifying density-based local outliers." ACM sigmod record. Vol. 29. No. 2. ACM, 2000. 12
Online Detection of Low-Quality PMU Data [6] [6] M. Wu, and Le Xie. Online Detection of Low-Quality Synchrophasor Measurements: A Data-Driven Approach. IEEE Transactions on Power Systems. Accepted, to appear. 13
Presentation Outline 1 Introduction 2 Technical Approach 3 Case Studies 4 Conclusions & Future Work 14
Numerical Results High Sensing Noise Test Case Description 39 real-world PMU voltage magnitude data curves. PMU No. 10, 15, 23, 29 contain Gaussian noises (SNR = 40 db) lasting from 1s to 1.2s. Line tripping fault is presented around 4s. Numerical Results Description All the 4 bad data segments are detected. System event does not cause false alarms. Detection delay is less than 0.19s. Computation time for each data window is 0.0376s. 15
Numerical Results Data Spikes Test Case Description 22 real-world PMU real power data curves. PMU No. 10, 13, 16, 21 contain data spikes lasting from 1.05s to 1.1s. Line tripping fault is presented around 4s. Numerical Results Description All the 4 bad data segments are detected. System event does not cause false alarms. Detection delay is less than 0.18s. Computation time for each data window is 0.0197s. 16
Numerical Results False Data Injection Attacks Test Case Description 39 real-world PMU voltage magnitude data curves. PMU No. 14, 18, 24, 37 contain false data injections lasting from 1s to 1.2s. Line tripping fault is presented around 4s. Numerical Results Description All the 4 false data injections are detected. System event does not cause false alarms. Detection delay is less than 0.19s. Computation time for each data window is 0.040s. 17
Numerical Results Un-updated Data Test Case Description 13 real-world PMU current magnitude data curves. PMU No. 1, 5, 7, 13 contain un-updated data lasting from 1s to 1.2s. Line tripping fault is presented around 4s. Numerical Results Description All the 4 bad data segments are detected. When Un-updated Data Is Presented System event does not cause false alarms. LOF Values (pu) 5 x 104 4 Detection delay is less than 0.18s. 3 Computation 2 time for each data window is 0.0115s. 1 0 LOF Values of Synchrophasor Channels LOF Threshold 1 2 3 4 5 6 7 8 9 10 11 12 13 Index of Synchrophasor Channels Current Magnitude (A) LOF Values (pu) LOF Values (pu) 1.05 x 104 1.04 1.03 1.02 Un-updated Data 1.01 0 1 2 3 4 5 6 7 8 Time (s) 5 x 104 4 3 2 1 0 100 50 0 Synchrophasor Measurements with Un-updated Data Synchrophasor Channel No. 1 Synchrophasor Channel No. 5 Synchrophasor Channel No. 7 Synchrophasor Channel No. 13 LOF Values of Synchrophasor Channels When Un-updated Data Is Presented LOF Threshold 1 2 3 4 5 6 7 8 9 10 11 12 13 Index of Synchrophasor Channels LOF Values of Synchrophasor Channels When Physical Event Is Presented LOF Threshold 1 2 3 4 5 6 7 8 9 10 11 12 13 Index of Synchrophasor Channels 18
Numerical Results Different Similarity Metrics Similarity Metric for High-Variance Bad Data Similarity Metric for Low-Variance Bad Data Performance Difference f H i, j is more sensitive to high-variance bad data. f L (i, j) is more sensitive to low-variance bad data. 19
Numerical Results Different Similarity Metrics LOF indicator based on f H (i, j) is more sensitive to high-variance bad data. LOF indicator based on f L i, j is more sensitive to low-variance bad data. 20
Presentation Outline 1 Introduction 2 Technical Approach 3 Case Studies 4 Conclusions & Future Work 21
Real-Time Detection of Low-Quality PMU Data Conclusions An approach for PMU low-quality data detection is proposed: It is purely data-driven, without involving any knowledge on network parameters or topology, which avoids the impact of incorrect parameter/topology information on the identification results. It encounters no convergence issues and has fast computation performance, which is desirable for online application. It is suitable for identifying low-quality data in PMU outputs under both normal and eventful operating conditions. Future Work Identify the root cause of the low-quality PMU data. Propose correction mechanism for the low-quality PMU data. 22
References [1] California ISO, Five year synchrophasor plan, California ISO, Tech. Rep., Nov 2011. [2] S. Ghiocel, J. Chow, et al. "Phasor-measurement-based state estimation for synchrophasor data quality improvement and power transfer interface monitoring," IEEE Tran. Power Systems, 2014. [3] K. D. Jones, A. Pal, and J. S. Thorp, Methodology for performing synchrophasor data conditioning and validation, IEEE Tran. Power Systems, May 2015. [4] M. Wang, J. Chow, P. Gao, X. Jiang, Y. Xia, S. Ghiocel, B. Fardanesh, G. Stefopolous, Y. Kokai, N. Saito, and M. Razanousky, A low-rank matrix approach for the analysis of large amounts of power system synchrophasor data, in System Sciences (HICSS), 2015 48th Hawaii International Conference on, Jan 2015, pp. 2637 2644. [5] K. Martin, Synchrophasor data diagnostics: detection & resolution of data problems for operations and analysis, in Electric Power Group Webinar Series, Jan 2014. [6] M. Wu, and Le Xie. Online Detection of Low-Quality Synchrophasor Measurements: A Data-Driven Approach. IEEE Transactions on Power Systems. Accepted, to appear. [7] Breunig, Markus M., et al. "LOF: identifying density-based local outliers." ACM sigmod record. Vol. 29. No. 2. ACM, 2000. 23
24